Lesson Objectives

  1. Determine stormflow and baseflow from high frequency flow data
  2. Use a hysteresis plot to understand watershed dynamics
  3. Communicate findings with peers through oral, visual, and written modes

Opening Discussion

High frequency data is usually defined as frequencies significantly lower than daily (e.g. 5-minute, 15-minute, 1 hr etc). What types of hydrological and biological processes happen on this timescale that we might want to investigate?

Session Set Up

getwd()
## [1] "C:/Users/Cathy/Box Sync/Hydrologic Data Analysis/Lessons"
packages <- c(
  "dataRetrieval", "tidyverse", "cowplot", "EcoHydRology", "xts", "dygraphs"
  )
invisible(lapply(packages, library, character.only = TRUE)) 

theme_set(theme_classic(base_size = 12))

Pull Data

dischargedatTFC2018 <- readNWISuv(
  site = "02097280", #Third Fork Creek at Woodcroft Parkway Near Blands
  parameterCd = "00060",
  startDate = "2018-01-01", 
  endDate = "2019-01-01"
  ) %>% 
  renameNWISColumns()


dischargedatEC2018 <- readNWISuv(
  site = "0208675010", #Ellerbe Creek at Club Boulevard at Durham, NC
  parameterCd = "00060", 
  startDate = "2018-01-01", 
  endDate = "2019-01-01"
  ) %>% 
  renameNWISColumns()


plot_grid(
  ggplot(dischargedatTFC2018, aes(x = dateTime, y = Flow_Inst)) +
    geom_line() + 
    scale_y_log10() +
    ggtitle("Third Fork Creek"),
  ggplot(dischargedatEC2018, aes(x = dateTime, y = Flow_Inst)) + 
    geom_line() + 
    scale_y_log10() +
    ggtitle("Ellerbe Creek"),
  ncol = 1
)

The shape and size of river hydrographs following rain events can tell us how water moves through a watershed. Once rain falls, how does water get to a stream? What would the streamflow do?

“Hydrograph separation is one of the most desperate analysis techniques in use in hydrology” - Hewlett & Hibbert 1967

TFCbaseflow <- BaseflowSeparation(
  dischargedatTFC2018$Flow_Inst, 
  filter_parameter = 0.925, 
  passes = 3
  )

TFC2018 <- cbind(dischargedatTFC2018, TFCbaseflow)

ggplot(TFC2018, aes(x = dateTime, y = Flow_Inst)) + 
  geom_line() +
  # scale_y_log10() +
  geom_line(mapping = aes(x = dateTime, y = bt), color = "darkorange4") +
  geom_line(mapping = aes(x = dateTime, y = qft), color = "steelblue4")

dygraph(
  cbind(
    Flow = with(TFC2018, xts(Flow_Inst, order.by = dateTime)), 
    Baseflow = with(TFC2018, xts(bt, order.by = dateTime)), 
    Quickflow = with(TFC2018, xts(qft, order.by = dateTime))
    )
  ) %>% 
  dyRangeSelector()
Export <- TFC2018 %>%
  mutate(timestep = c(diff(as.numeric(dateTime)), NA_real_),
         baseflowexport = bt * timestep,
         quickflowexport = qft * timestep) %>%
  summarize(BaseflowExport_cf = sum(baseflowexport, na.rm = T),
            QuickflowExport_cf = sum(quickflowexport, na.rm = T),
            TotalExport_cf = BaseflowExport_cf + QuickflowExport_cf)

What percentage of water is exported from this watershed as baseflow? What percentage as quickflow?

Repeat this analysis for Ellerbe Creek. How do the percentages compare?

What does this mean physically about the watershed?

Hydrologic Flashiness

A “flashy” stream increases in flow faster following a precipitation event.

Refering to our previous work with baseflow separation, would you expect flashiness to increase or decrease with urbanization?

What do you think “urban stream syndrom” refers to?

The Richards-Baker Index is an index of how flashy a stream is. It is calculated as \[RBI = (\sum_{j = 1}^n |Q_j - Q_{j-1}| / \sum_{j = 1}^n Q_j) / Catchment Area\]

TFCsitedata <- readNWISsite(site = "02097280")
TFC.catchment.size <- TFCsitedata$drain_area_va #square miles

RBI_TFC2018 <- (
  sum(abs(diff(dischargedatTFC2018$Flow_Inst))) / 
    sum(dischargedatTFC2018$Flow_Inst[-1])
  ) / TFC.catchment.size

ECsitedata <- readNWISsite(site = "0208675010")
EC.catchment.size <- ECsitedata$drain_area_va #square miles

RBI_EC2018 <- (
  sum(abs(diff(dischargedatEC2018$Flow_Inst))) / 
    sum(dischargedatEC2018$Flow_Inst[-1])
  ) / EC.catchment.size

Which stream is flashier?

Chemical Flashiness

Chemical concentrations in rivers can be either highly variable with discharge, or “chemostatic”, which means the concentration barely changes with discharge.

CCdat <- readNWISuv(
  site = "02249500", # CRANE CREEK AT MELBOURNE, FL
  parameterCd = c("00060", "99133"), # Discharge in cfs & Nitrate in mg/l NO3-N
  startDate = "2015-01-1",
  endDate = "2016-01-01"
  ) %>%
  renameNWISColumns() %>%
  rename(Nitrate_mgl = 6)

ggplot(CCdat,
       aes(x = Flow_Inst, y = Nitrate_mgl)) +
  geom_point() +
  scale_x_log10() +
  scale_y_log10()
## Warning: Removed 4434 rows containing missing values (geom_point).

Potdat <- readNWISuv(
  site = "01646500", # POTOMAC RIVER NEAR WASH, DC LITTLE FALLS PUMP STA
  parameterCd = c("00060", "99133"), # Discharge in cfs & Nitrate in mg/l NO3-N
  startDate = "2016-01-1",
  endDate = "2017-01-01"
  ) %>% 
  renameNWISColumns() %>%
  rename(Nitrate_mgl = 6)

ggplot(Potdat,
       aes(x = Flow_Inst, y = Nitrate_mgl)) +
  geom_point() +
  scale_x_log10() +
  scale_y_log10()
## Warning: Removed 3564 rows containing missing values (geom_point).

Which of these rivers is more chemostatic?

Hysteresis Loops

Hysteresis occurs when the concentration varies with discharge in a different relationship on the rising and falling limbs of a storm. This can be caused by flushing behavior or dilution behavior.

dygraph(
  cbind(
    Flow = xts(Potdat$Flow_Inst, order.by = Potdat$dateTime), 
    Nitrate = xts(Potdat$Nitrate_mgl, order.by = Potdat$dateTime)
    )
  ) %>% 
  dySeries("Nitrate", axis = "y2") %>%
  dyRangeSelector()

Look at the storm that occured around February 1st. Do you expect this storm to be a flushing or a diluting storm?

PotomacStorm <- Potdat %>%
  filter(dateTime > "2016-01-31" & dateTime < "2016-02-14") 

ggplot(PotomacStorm, aes(x = Flow_Inst, y = Nitrate_mgl, color = dateTime)) +
  geom_point()

Did this storm exhibit clockwise or counterclockwise hysteresis?

Which direction do hysteresis loops go for diluting storms?

Which direction do hysteresis loops go for flushing storms?

Closing Discussion

What hydrological processes did we investigate today?

What can high-frequency data be used for that daily data may not be able to capture?